Everyone has a theory of why polls are sometimes incorrect, from poor methodologies to unlucky, unrepresentative samplings. But then there are the polls where one simple explanation is probably not enough. Case in point: a survey done by Mitchell Research & Communications of Michigan Democratic primary voters a mere two days before Tuesday night's primary. What did it find? Well, according to the poll Hillary Clinton led Bernie Sanders by the oh-so-narrow margin of...37 percent. That's right, Clinton bested Sanders in the poll 66 percent to 29 percent, which is not far from the margin by which Bashar al-Assad wins elections.
Well, the results from Michigan came in tonight and the race was closer than expected. Clinton didn't win by 37 percent; indeed, she only won by...-2 percent. That's a 39 percent point swing, and a victory for Bernie Sanders. It's of course possible some fluke event occurred between the poll's release and the election. Perhaps, as has been speculated, Democratic voters, sensing a foregone conclusion, voted in the Republican race, thanks to Michigan's open primary. Perhaps Sanders's debate performance Sunday night was more popular with voters than commentators. Or perhaps the sample was just screwy. Whatever the reason, this has to go down as one of the least accurate polls taken within 48 hours of a statewide election.